Online platforms provide a large number of products/information for consumers to browse. It brings convenience for gathering information, however, a problem appears when consumers make decisions. Since the products on platforms have multiple attribute levels, it is complex for consumers to compare products with considering all these attribute levels. In this case, consumers need a heuristic decision rule to facilitate the decision making. For example, a consumer can use the filtering options provided by online intermediaries to directly find the products with some most important attribute levels based on preference. The decision rule that the consumer uses during the filtering is a non-compensatory rule, which is called the lexicographic rule. The selection for sequential search...[ Read more ]

Online platforms provide a large number of products/information for consumers to browse. It brings convenience for gathering information, however, a problem appears when consumers make decisions. Since the products on platforms have multiple attribute levels, it is complex for consumers to compare products with considering all these attribute levels. In this case, consumers need a heuristic decision rule to facilitate the decision making. For example, a consumer can use the filtering options provided by online intermediaries to directly find the products with some most important attribute levels based on preference. The decision rule that the consumer uses during the filtering is a non-compensatory rule, which is called the lexicographic rule. The selection for sequential search behaviors is suitable for the situation when consumers apply the lexicographic rule in the previous literature. Also, since the selection of the filtering keywords has not been explored in the previous literature, in this study, we proposes a sequential search model with Dirichlet learning to investigate how consumers choose the search behaviors, including the filtering keywords, typically assuming that they use lexicographic rules to make the search decisions. Our approach has three distinct advantages. First, we endogenize a consumer’s response to the default ranking with search refinement tools. By modeling the selection of filtering keywords, the model can estimate the consumers search actions efficiently. Second, by estimating the lexicographic sequences used by consumers, researchers can directly find the differences of consumer preferences. Third, this study could contribute to the online platform design by providing appropriate filtering options for consumers to improve their search efficiency. Next, we simulate a data set and estimate the parameters of the model. The results show that our method could precisely replicate a consumer’s choices and the consumer’s lexicographic preference.